Only save combined data_type in only-combining mode (#878)

* Only save combined `data_type` in only-combining mode

* Retrigger docs rebuild

* Retrigger docs rebuild

* Add more tests

strax/context.py

@@ -1274,7 +1274,11 @@ def concat_loader(*args, **kwargs):
                 # only save if we are not in a superrun or the plugin allows superruns
                 # otherwise we will see error at Chunk.concatenate
                 # but anyway the data is should already been made
-                for d_to_save in set(current_plugin_to_savers + list(target_plugin.provides)):
+                if not _combining_subruns:
+                    data_type_to_save = set(current_plugin_to_savers + list(target_plugin.provides))
+                else:
+                    data_type_to_save = set(current_plugin_to_savers)
+                for d_to_save in data_type_to_save:
                     key = self.key_for(run_id, d_to_save, chunk_number=chunk_number)
                     # Here we just check the availability of key,
                     # chunk_number for _get_partial_loader_for can be None
@@ -1333,7 +1337,7 @@ def concat_loader(*args, **kwargs):
             targets=strax.to_str_tuple(final_plugin),
         )
 
-    def get_datakey(self, run_id, target, lineage):
+    def get_data_key(self, run_id, target, lineage):
         """Get datakey for a given run_id, target and lineage.
 
         If super is detected, the subruns information are added to the key.
@@ -1366,7 +1370,7 @@ def _add_saver(
         :return: Updated savers dictionary.
 
         """
-        key = self.get_datakey(run_id, d_to_save, target_plugin.lineage)
+        key = self.get_data_key(run_id, d_to_save, target_plugin.lineage)
         for sf in self._sorted_storage:
             if sf.readonly:
                 continue
@@ -2020,7 +2024,7 @@ def key_for(self, run_id, target, chunk_number=None):
             plugins = self._get_plugins((target,), run_id, chunk_number=chunk_number)
 
         lineage = plugins[target].lineage
-        return self.get_datakey(run_id, target, lineage)
+        return self.get_data_key(run_id, target, lineage)
 
     def get_meta(self, run_id, target, chunk_number=None) -> dict:
         """Return metadata for target for run_id, or raise DataNotAvailable if data is not yet

strax/run_selection.py

@@ -68,7 +68,7 @@ def keys_for_runs(
         # Get the lineage once, for the context specifies that the
         # defaults may not change!
         p = self._get_plugins((target,), run_ids[0])[target]
-        return [self.get_datakey(r, target, p.lineage) for r in run_ids]
+        return [self.get_data_key(r, target, p.lineage) for r in run_ids]
     else:
         return []
 

tests/test_core.py

@@ -447,6 +447,20 @@ def test_per_chunk_storage():
             register=[Records, Peaks],
             use_per_run_defaults=True,
         )
+
+        # If per-chunk storage is not for dependencies, DataKey will not be different.
+        key_1st = st.key_for(run_id, "records", chunk_number={"records": [0]})
+        key_2nd = st.key_for(run_id, "records", chunk_number={"records": [1]})
+        assert str(key_1st) == str(key_2nd)
+
+        # If per-chunk storage is for dependencies, savers will not be different.
+        components_1st = st.get_components(run_id, "peaks", chunk_number={"peaks": [0]})
+        components_2nd = st.get_components(run_id, "peaks", chunk_number={"peaks": [1]})
+        assert (
+            components_1st.savers["peaks"][0].dirname == components_2nd.savers["peaks"][0].dirname
+        )
+
+        # Test merge_per_chunk_storage
         st.make(run_id, "records")
         n_chunks = len(st.get_metadata(run_id, "records")["chunks"])
         assert n_chunks > 2
@@ -462,6 +476,7 @@ def test_per_chunk_storage():
         with pytest.raises(ValueError):
             st.merge_per_chunk_storage(run_id, "peaks", "records")
 
+        # Per-chunk storage not allowed for some plugins
         p = type("whatever", (strax.OverlapWindowPlugin,), dict(depends_on="records"))
         st.register(p)
         with pytest.raises(ValueError):

tests/test_superruns.py

@@ -126,7 +126,9 @@ def test_loaders_and_savers(self):
             self.superrun_name, "peak_classification", _combining_subruns=True
         )
         assert len(components.loaders) == 1
+        assert len(components.savers) == 1
         assert "peak_classification" in components.loaders
+        assert "peak_classification" in components.savers
 
         with self.assertRaises(ValueError):
             self.context.get_components(
@@ -335,6 +337,8 @@ def test_only_combining_superruns(self):
         The test also shows the difference between the two.
 
         """
+        self.context.tree
+        self.context.inversed_tree
         self.context.check_superrun()
         sum_super = self.context.get_array(self.superrun_name, "sum")
         _sum_super = self.context.get_array(self.superrun_name, "sum", _combining_subruns=True)